Aaron Byrne

AI · Quantitative Modelling · Data Science · Finance

I build intelligent systems where human ingenuity, AI, and finance converge.

I'm interested in the foundations of trustworthy intelligence: how systems represent knowledge, reason under uncertainty, and turn complex information into reliable decisions.

Currently

I'm working on Claim-Driven Development (CDD) , a new software development paradigm built around mathematically checkable claims rather than source code, as the primary source of truth. I'm now actively building the first CDD conforming implementation (focusing initially on python). This is being designed for AI-assisted code development where creating code is cheap, but understanding and verifying it, matters more than ever.

Recent writing

All writing
  • The True Cost of Separating Code Development from Code Understanding in an AI first world

    2026-08-13

    When AI makes code cheap, understanding becomes expensive*

  • The number that broke arithmetic

    2026-08-07

    The Greeks proved a number existed that they could not write down, and it wrecked their programme. The repair took 2,200 years and is the single property every convergence proof in this series quietly uses. Then the modern version of the same problem: why 0.1 + 0.2 is not 0.3, and why the textbook variance formula should not go into production.

  • The map in one hour

    2026-08-05

    Hilbert space, manifold, measure, homology, transport plan. One sentence each, one paragraph each, and what each one lets you do that you could not do before. No proofs and no prerequisites: this is the map, so that everything after it has somewhere to attach.